Confounders In Casual Inference And Bayesian Networks

This paper mainly discusses the estimation precision of the relative risk in casual inference, parameter estimation of the distribution with non-ignorable missing data in longitudinal studies, an introduction for Bayesian networks and MMHC in Casual Explorer. It consists of the following three parts:(1) For a non-confounder, it is argued by many authors about whether stan-dardization can improve the precision of estimation. This paper discusses the estimation precision of the relative risk in causal inference. It is shown that the estimation precision of relative risk cannot be improved by standardizing for a non-confounder no matter how to re-categorize the non-confounder. We finally developed four asymptotic interval estimators for RR (relative risk) in a stratified population, and we gave a proposing simulation study.(2) It is not uncommon to encounter missing data can not be ignored in longitudinal studies. In this paper, we focus attention on parameter estimation of the distribution with missing data. We prove that the distribution of pa-rameters can be expressed by observed data. We also consider the relationship between variance of the parameter estimation and instrumental variable ORY1Y2 , Applying simulation, we find there is a positive correlation between them. A comparing simulation is also presented to confirm our conclusion.(3) A Bayesian network structure learning is discussed, After that, I empha-sized MMHC in Casual Explorer. At last, it gets good,performance by applying the economic data.